Human Activity Recognition Based on a Modified Capsule Network

Table 9

A comparison of the proposed approach with other studies on UCI-HAR.

Dataset	Model	Accuracy (%)
UCI-HAR	CNN [49]	92.50
	Attention-based CNN [50]	93.16
	iSPLInception [51]	95.09
	Ensem-HAR [52]	95.05
	CNN-GRU [53]	94.50
	Fusing 2-D FFT and WVT [54]	93.45
	Deep CNN-LSTM with self-attention [55]	93.11
	CNN-BAOA [56]	95.23
	MCN (this paper)	95.28